Microbiome gut community structure and functionality are associated with symptom severity in non-responsive celiac disease patients undergoing a gluten-free diet

Abstract

Non-responsive celiac disease (NRCD) challenges clinicians due to persistent symptoms despite a gluten-free diet (GFD). We present a cross-sectional pilot study including 39 NRCD patients to describe the underlying mechanisms contributing to symptom persistence by integrating different levels of data (fecal shotgun metagenomics, mucosal integrity markers, and metabolomic profiles) and using microbial networks to unravel the community structure of the patient's microbiome. Two distinct clusters of patients were identified based on clinical and demographic variables not influenced by gluten consumption. Cluster 1, labeled "Low-grade symptoms," displayed milder symptoms and lower inflammatory markers and a fragmented microbial network characterized by high modularity and a reliance on localized hubs, suggesting a microbial community under stress but capable of maintaining limited functionality. Cluster 2, named "High-grade symptoms," exhibited more severe symptoms, elevated inflammatory markers, and a more connected but antagonistic microbial network with a greater number of keystone taxa, including taxa associated with Th17 activation and inflammation. In contrast, the control network, representing asymptomatic treated celiac disease (tCD) patients, was highly interconnected, resilient, and cooperative, with a robust structure maintained even under simulated disruptions. Metabolomic analysis revealed differential metabolites between clusters, particularly those involved in amino acid metabolism pathways and microbial-derived metabolites such as indolelactic acid and mannitol, which were associated with symptom severity. This study identifies NRCD subgroups based on the gut microbiome and metabolic signatures associated with clinical manifestations, highlighting variations in microbial network stability and metabolite profiles as contributors to symptom persistence and potential therapeutic targets.

Importance: Celiac disease (CD) is a chronic immune-mediated systemic disorder caused by consuming gluten in genetically susceptible individuals. There is currently no cure for CD, and the most effective treatment is maintaining a strict, lifelong gluten-free diet (GFD). This nutritional therapy aims to prevent the immune reaction triggered by gluten and promote the healing of the intestinal lining, resolving the clinical, serological, and histological abnormalities within 6-12 months. However, up to 30% of patients may continue to experience symptoms or exhibit laboratory abnormalities or intestinal inflammation suggestive of active CD, despite following a GFD. This challenge, which encompasses various diagnoses, is known as nonresponsive celiac disease (NRCD). In this study, we explored the role of intestinal microbiota in causing NRCD, finding an association between the persistence of symptoms and changes in mucosal integrity biomarkers, with different gut microbiome structures among NRCD patients, indicating a significant role of the microbiome in NRCD.

Keywords: celiac disease; co-occurrence networks; gluten-free diet; metabolome; microbiome; symptom association.

Fig 1

Microbial community structure analysis: (A) Alpha diversity measured by Chao1, Shannon, and Simpson diversity indexes. (B) Multidimensional scaling (MDS) of the Unifrac distance representing beta diversity of the samples. (C) Cladogram showing the taxonomic relations of the microbial markers identified after linear discriminant analysis effect size (LEfSe) analysis using an LDA score >3. (D) MetaCyc pathways differing significantly between NRCD patients with low-grade symptomatology (low-NRCD) and high-grade symptomatology (high-NRCD) with an effect size ≥0.75%.

Fig 2

Microbial network analysis: Co-occurrence networks of the microbiome from (A) asymptomatic treated CD patients (tCD). (B) NRCD patients with low-grade symptoms (low-NRCD). (C) NRCD patients with high-grade symptoms (high-NRCD). (D) Network of differentially associated species in NRCD patients with low-grade symptoms (low-NRCD). (E) Network of differentially associated species in NRCD patients with high-grade symptoms (high-NRCD). (F) Number of nodes in the networks’ LCC compared to the original after sequential removal of nodes with the highest degree. (G) Same as panel F, but with removal of nodes based on their highest betweenness centrality. (H) Same as panel F, but with removal of random nodes. The nodes in the graph represent different bacterial species. The size of each node corresponds to its relative abundance, while the edges represent statistically significant associations between the nodes (P < 0.05). Green edges indicate positive relationships, while red edges indicate negative ones. The thickness of each edge indicates the strength of the association. Nodes highlighted in panels A, B, and C correspond to keystone taxa calculated as nodes with betweenness centrality and degree greater than log-normal quantile 0.90.

Fig 3

(A) Metabolite pathway enrichment analysis (MPEA) of microbiota-derived (bacterial) and co-metabolism-derived metabolites. (B) Number of pathways of bacterial and host co-metabolism origin. Bar plot showing the relative significance of differential metabolic pathways from MPEA according to the differential metabolites from bacterial and host co-metabolism origins.

Fig 4

Network integrating microbiome and metabolome interactions. Differential metabolites from microbiota and co-metabolism origins and their related bacteria and involved KEGG metabolic pathways are shown. Nodes represent bacterial species, metabolites, or pathways. The three kinds of nodes are differentiated by node shape. Node size is proportional to node degree. Microbe and metabolite nodes are colored depending on the cluster in which they are upregulated. Pathway node color represents whether it is present only in microbial or also in host metabolism. Edges between microbes and metabolites are colored based on the correlation sign. Edges representing significant correlations (P < 0.05) are represented by a thicker line.

Fig 5

Heatmap representation of Pearson’s correlation between changes in symptom scores and changes in gut microbiome (keystone and differential associated taxa), microbial functional potential (differential abundance metaCyc pathways), and fecal metabolome (differentially abundant metabolites). Cluster annotation represents if the value is more abundant in the low-NRCD group (pink) or high-NRCD group (brown). CeD-Pro: Celiac Disease Patient-Reported Outcome. GSRS: Gastrointestinal Symptom Rating Scale Celiac Disease. Significant results: (*) P < 0.05; (**) P < 0.01.